Nonmetric multivariate analysis in community-level ecotoxicology

Community-leveI data, typically in the form of abundances of over 100 speci es, are widely collected in the context of environmental monitoring, e.g., of the effects of disposal of drilling muds in offshore oil operations. The statistical properties of the resulting abundance arrays preclude the use of "classical" multivariate analyses, such as principal components and mult ivariate analysis of variance. One alternative is the use of nonparametric displays and tests, such as nonmetric multidimensional scaling (MDS) and va riations of Mantel rests on similarity matrices. These do not require the r estrictive assumptions of parametric techniques and possess a conceptual si mplicity, facilitating their use and understanding by environmental manager s and regulators. A monitoring example is discussed from Norwegian oil fiel ds, for which such analyses have had a significant impact on environmental practice. The techniques are equally applicable to assessing outcomes from community-level laboratory experiments and bioassays. The paper exemplifies the analysis approach taken at the Plymouth Marine Laboratory (and encapsu lated in the PRIMER software) through (1) observational studies of pollutio n gradients in North Sea sediments and heavy metal pollutants in the Fal es tuary, UK; (2) an experimental study on differential effects of metals on m arine nematode communities; and (3) a bioassay approach employing a microco sm experiment on Fal estuary sediments.